74LS644 availability

Hello, I am working on my own ISA card for storage and peripheral interface:


I decided against using the 74LS644 because I thought a new design in 2022 should not use a chip that is already rare, so I improvised the function from XOR 74LS86 and OR 74LS32 gates. Is it worth it to avoid such specialized, old chips for new designs even if that means it makes the board a little more complicated? I can find Chinese companies making clones of simple logic chips, but not complicated logic operations like the '688. I have a current picture of the card below.

Dan

profdc9 said:

I decided against using the 74LS644 because I thought a new design in 2022 should not use a chip

Click to expand...

Just to be clear, the thread title and this sentence is a typo, right? The '644 is basically an oddball variant of the '245 octal buffer, not a comparator, so I was scratching my head why in the world you'd have wanted to use it and how OR gates were a replacement for it.

The 688 is still made in the 74HC/T and 74LS/ALS logic families and are still pretty readily available from places like DigiKey, so I wouldn't really fret too much using it. It'll probably stick around for as long as HC/T and LS/ALS logic in general does, honestly, it's a pretty popular legacy chip.

The 688 is neither rare nor unavailable. You can even find 521s around (Fairchild equivalent).

At any rate, if you want to avoid the 688 for equality, you can use two 74LS85s and a single 2-input AND gate on the "=" outputs. You could also program a GAL to do the same function. If you don't need the tristate control on the 688, you could also use a 684. Lots of ways to skin this cat.

Chuck(G) said:

You could also program a GAL to do the same function

Click to expand...

This is my favorite go-to; an atf16v8 or 22v10 doesn't cost much more than a '688 and instead of being limited to 8 address lines with the latter you could completely decode a 20 bit address if that's what you really needed. (Or combine the decode with a bunch of other moving parts.) Only downside is the current chip shortage has really nailed the supply chain for the new ones.

(That said I've had decent luck with used GALs from AliExpress and other usual suspects. Haven't gotten a "fake" one yet, and they're designed to be re-flashed.)

I am just worried with the semiconductor shortage many of these chips may never be made again once current stocks run out. If a Chinese company is cloning it, perhaps that means there will be other sources of it for awhile. I've been burned a few times by parts that are no longer supported, and if I am making an open source project, it would be nice if when it was being designed, it was designed for parts that at the time of the design were anticipated to be available for a long time to come.

profdc9 said:

I am just worried with the semiconductor shortage many of these chips may never be made again once current stocks run out.

Click to expand...

If this happens it means that 74 series logic is extinct, so it’s all sort of moot.

FWIW, your board uses an Atmega MCU in DIP format, have you seen the expected delivery dates for new batches of those? I’m personally a lot more worried about not being able to get those than ‘688s.

Just out of curiosity, where are you finding catalogs for ”knockoff” Chinese chip suppliers to know definitively that none of them are making ‘688s, with or without a Western corporate logo silkscreened on them?

I was looking at lcsc.com because they tend to have parts from many Chinese suppliers, and I have found them pretty reliable in representing sources of ICs. I was encouraged to use 74LS86/74HCT86 and 74LS32/74HCT32 because they are so fundamental and there's three to four manufacturers of those. There's an interesting manufacturer XINLUDA that makes TTL logic currently that is accessible on lcsc.com and could be the source of many generic logic ICs. I do not know who is making the 74LS688 I find on ebay and aliexpress, and I can only find Texas Instruments as providing this chip and many other TTL/HCT chips, which is somewhat worrisome I think. I don't want to make a new design now for which there is only one manufacturer of chips, and TI seems to be the principal manufacturer of many older TTL chips.

I know the lead time on ATMEGA328P is long, but it seems unlikely it will be obsolete soon given Microchip is manufacturing it and it is so widespread because of the Arduino. Also, there is a clone from Logic Green called the LGT32F328P. There are a few differences between these two (higher clock speed, single cycle execution, how they're programmed, and in some of the peripherals) but they are very similar. It would also be relatively easy to layout the board for the LQFP-32 ATMEGA328P if needed.

I have also made a SPI interface for SD card access from TTL/CMOS components for my Z80 computer design http://www.github.com/profdc9/Z80SBC which perhaps I should use instead. But including the microcontroller increases the flexibility a lot.

TI is effectively just a label. All of their chips, I guarantee you, are churned out on the same foundry lines in China as everything else. These chip designs are all way past any patent ownership, so if for some reason TI corporate decides to take it off the list any one of these foundaries can just keep churning them out with another label on them tomorrow. As Chuck said, I see no current shortage of ‘688s (while Atmegas are actually a little hard to find right now) so, *shrug*, it just seems an odd thing to pick on.

... Especially if you’re fine with, if you have to, just re-doing the board layout if/when the situation calls for it. “Xinluda” does make 74ls85s, so if somehow they magically end up the only foundary and brand left in the world making 74 logic and they stubbornly refuse to churn out 688s (the fact that anyone is still making them certainly suggests there does exist a demand) you could port a ‘688 design to a pair of ‘85s trivially.

Nothing we do now is guarunteed to relevant in 40 years. Or tomorrow.

As Eudimorphodon observed, the 688 isn't going away anytime soon. As I observed, using some LS85s and one section of an LS00 will save on BOM count.

The chip shortage is mostly hitting the "older" chips, mostly due to the automotive sector's demand (because of safety certification requirements, you don't use "bleeding edge" parts). The other reason is that fabs have been converting over to 300 mm wafer gear and high-end products. The old parts lines use fab equipment for 200 mm wafers--and nobody's building those anymore.

Chuck(G) said:

one section of an LS00

Click to expand...

You could even use one of those cute 1gate parts instead of a whole ‘00 if board space outweighs the hassle of tacking down a surface mount tic-tac.

Al Kossow said:

Absolutely true, and TI, the company everyone loves to hate, is EOL a LOT of parts

Click to expand...

If you look up the CD74HCT688 on Ti’s website their February 2022 update to the datasheet (and the search engine html) declares it an ACTIVE part “recommended for new designs”. This “Active” designation also applies to the HC/LS/ALS flavors.

In theory at least there’s actually two stages of labels a chip is supposed to go through before TI stops making it, so maybe it’s better not to catastrophize about them killing this particular part until there’s some evidence it’s on the chopping block.

I regret to report a brain bubble. The 74LS85 is cascadable, through the three "I" inputs. No need for a separate gate. If you want to do a 24 bit magnitude comparator, you need only use 6 x 74LS85. Gotta read those datasheets!

I guess another thought just crystalized in my head: who's still making 82C55s? I guess Renesas claims to, but they're all out of stock and are listing one of those "try again in a year" delivery dates. Ohwell.

Dunno if any are in current production, but Alibaba has a lot of vendors pushing the NEC D71055 (82C55 equivalent). They're actually better than the Intel original.

I have tried the Xinluda 82C55 and they seem to work ok. It required a slight slow down in the Arduino, but otherwise works fine.

Heck, a lot of plain LSTTL requires a slowdown on the STM32407 MPUs. It can toggle I/Os at 84 MHz, which is way too fast for garden-variety TTL latches. I discovered that the hard way.